Igt-dev Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Louis Chauvet <louis.chauvet@bootlin.com>
To: Mark Yacoub <markyacoub@google.com>, igt-dev@lists.freedesktop.org
Subject: Re: [PATCH v2] tests/unigraf: Enhance link rate support checking and hardware retrain timing
Date: Fri, 31 Jul 2026 19:48:30 +0200	[thread overview]
Message-ID: <dfd7fa67-6120-4e60-babb-f454e181f6f7@bootlin.com> (raw)
In-Reply-To: <20260720205355.321557-1-markyacoub@google.com>



On 7/20/26 22:53, Mark Yacoub wrote:
> Refactor hardware capability checks to natively skip unsupported DP link parameters dynamically, moving validation cleanly into the core igt_dp abstraction layer robustly.
> 
> v2:
>   - Downgrade `igt_assert_f` checks across the `igt_dp` layer to `igt_require_f` (Louis)
>   - Extract link rate querying into generic `igt_dp_is_link_rate_supported` helper (Louis)
>   - Add `expected_new_connector_count` to `kms_wait_for_new_connectors` to fix discovery race conditions natively (Louis)
> 
> TAG=agy
> CONV=35e8e1e4-2dfd-4bf2-8c0b-c9bbc6268d92
> ---
>   lib/igt_dp.c                         | 50 ++++++++++++++++++++++++++--
>   lib/igt_dp.h                         |  1 +
>   lib/igt_kms.c                        | 15 ++++++---
>   lib/igt_kms.h                        |  4 ++-
>   lib/vendor/unigraf/unigraf.c         |  1 +
>   tests/unigraf/unigraf_connectivity.c | 19 +++++++++--
>   tests/unigraf/unigraf_lt.c           | 31 +++++++++++++----
>   7 files changed, 104 insertions(+), 17 deletions(-)
> 
> diff --git a/lib/igt_dp.c b/lib/igt_dp.c
> index 9f71c9e72..20af6a967 100644
> --- a/lib/igt_dp.c
> +++ b/lib/igt_dp.c
> @@ -107,7 +107,7 @@ int igt_dp_get_max_lane_count(int drm_fd, igt_output_t *output)
>   	if (is_intel_device(drm_fd))
>   		return i915_dp_get_max_lane_count(drm_fd, output);
>   
> -	igt_assert_f(false, "Current drm device is not able to report max lane count\n");
> +	igt_require_f(false, "Current drm device is not able to report max lane count\n");
>   	return -EINVAL;
>   }
>   
> @@ -124,7 +124,51 @@ void igt_dp_force_link_retrain(int drm_fd, igt_output_t *output, int retrain_cou
>   		return;
>   	}
>   
> -	igt_assert_f(false, "Current drm device does not support link retraining\n");
> +	igt_require_f(false, "Current drm device does not support link retraining\n");
> +}
> +
> +/**
> + * igt_dp_is_link_rate_supported: Tests if the specific link rate can be forced
> + * @drm_fd:		DRM file descriptor
> + * @output:		igt_output_t object representing the display port
> + * @rate_kb:		Link rate in kbps to test
> + *
> + * Returns: true if device supports enforcing this rate, false otherwise
> + */
> +bool igt_dp_is_link_rate_supported(int drm_fd, igt_output_t *output, int rate_kb)
> +{
> +	if (is_intel_device(drm_fd)) {
> +		char buf[512];
> +		int res;
> +		char *token;
> +		int rate_tens = rate_kb / 10;
> +
> +		res = igt_debugfs_read_connector_file(drm_fd, igt_output_name(output),
> +						      "i915_dp_force_link_rate",
> +						      buf, sizeof(buf));
> +		if (res < 0)
> +			return false;
> +
> +		token = strtok(buf, " \n\t,:");
> +		while (token) {
> +			int rate;
> +			char *endptr;
> +
> +			errno = 0;
> +			rate = strtol(token, &endptr, 0);
> +			/*
> +			 * Robustly verify that a number was actually parsed (endptr != token)
> +			 * allowing us to safely ignore formatting headers (e.g. "Rates:")
> +			 */
> +			if (!errno && endptr != token && rate == rate_tens)
> +				return true;
> +			token = strtok(NULL, " \n\t,:");
> +		}
> +
> +		return false;
> +	}
> +
> +	return igt_dp_get_max_supported_rate(drm_fd, output) >= rate_kb;
>   }

That a bit strange to copy/paste the existing 
i915_dp_get_max_supported_rate code here, did I miss something in your 
implementation? Or is the existing implementation broken?

>   
>   /**
> @@ -140,7 +184,7 @@ int igt_dp_get_pending_retrain(int drm_fd, igt_output_t *output)
>   	if (is_intel_device(drm_fd))
>   		return i915_dp_get_pending_retrain(drm_fd, output);
>   
> -	igt_assert_f(false, "Current drm device does not support pending retrain count checking\n");
> +	igt_require_f(false, "Current drm device does not support pending retrain count checking\n");
>   	return -EINVAL;
>   }
>   
> diff --git a/lib/igt_dp.h b/lib/igt_dp.h
> index 0c56ad2ec..e5cacad66 100644
> --- a/lib/igt_dp.h
> +++ b/lib/igt_dp.h
> @@ -22,6 +22,7 @@ int igt_dp_get_max_supported_rate(int drm_fd, igt_output_t *output);
>   int igt_dp_get_max_lane_count(int drm_fd, igt_output_t *output);
>   
>   void igt_dp_force_link_retrain(int drm_fd, igt_output_t *output, int retrain_count);
> +bool igt_dp_is_link_rate_supported(int drm_fd, igt_output_t *output, int rate_kb);
>   
>   int igt_dp_get_pending_retrain(int drm_fd, igt_output_t *output);
>   
> diff --git a/lib/igt_kms.c b/lib/igt_kms.c
> index 2eefb773b..041a3356b 100644
> --- a/lib/igt_kms.c
> +++ b/lib/igt_kms.c
> @@ -8255,10 +8255,17 @@ get_array_diff(const uint32_t *array_a, int array_a_len, const uint32_t *array_b
>    */
>   int kms_wait_for_new_connectors(uint32_t **newly_connected,
>   				const uint32_t *already_connected,
> -				int already_connected_count, int drm_fd)
> +				int already_connected_count,
> +				int expected_new_connector_count,
> +				int drm_fd)
>   {
>   	int newly_connected_count;
>   	struct timespec start, end;
> +	int timeout = igt_default_display_detect_timeout();
> +
> +	/* MST discovery may be serialized in the kernel; allocate extra time for topology settling */
> +	if (expected_new_connector_count > 1)
> +		timeout *= 2;
>   
>   	igt_assert(newly_connected);
>   	igt_assert_fd(drm_fd);
> @@ -8269,10 +8276,10 @@ int kms_wait_for_new_connectors(uint32_t **newly_connected,
>   			free(*newly_connected);
>   		newly_connected_count = igt_get_connected_connectors(drm_fd, newly_connected);
>   		igt_assert_eq(igt_gettime(&end), 0);
> -	} while (!get_array_diff(*newly_connected, newly_connected_count,
> +	} while (get_array_diff(*newly_connected, newly_connected_count,
>   			       already_connected, already_connected_count,
> -			       NULL) &&
> -		 igt_time_elapsed(&start, &end) <= igt_default_display_detect_timeout());
> +			       NULL) < expected_new_connector_count &&
> +		 igt_time_elapsed(&start, &end) <= timeout);
>   
>   	return newly_connected_count;
>   }
> diff --git a/lib/igt_kms.h b/lib/igt_kms.h
> index 521a03c01..09cec556f 100644
> --- a/lib/igt_kms.h
> +++ b/lib/igt_kms.h
> @@ -1382,7 +1382,9 @@ uint32_t igt_get_connector_id_from_mst_path(int drm_fd, const void *mst_path);
>   
>   int kms_wait_for_new_connectors(uint32_t **newly_connected,
>   				const uint32_t *already_connected,
> -				int already_connected_count, int drm_fd);
> +				int already_connected_count,
> +				int expected_new_connector_count,
> +				int drm_fd);
>   int
>   get_array_diff(const uint32_t *array_a, int array_a_len, const uint32_t *array_b, int array_b_len,
>   	       uint32_t **diff);
> diff --git a/lib/vendor/unigraf/unigraf.c b/lib/vendor/unigraf/unigraf.c
> index 30ee3c72b..bfb06bd0c 100644
> --- a/lib/vendor/unigraf/unigraf.c
> +++ b/lib/vendor/unigraf/unigraf.c
> @@ -270,6 +270,7 @@ static void unigraf_autodetect_connector(int drm_fd)
>   	newly_connected_count = kms_wait_for_new_connectors(&newly_connected,
>   							    already_connected,
>   							    already_connected_count,
> +							    1,
>   							    drm_fd);
>   
>   	diff_len = get_array_diff(newly_connected, newly_connected_count,
> diff --git a/tests/unigraf/unigraf_connectivity.c b/tests/unigraf/unigraf_connectivity.c
> index 8ae9a2593..e813c9973 100644
> --- a/tests/unigraf/unigraf_connectivity.c
> +++ b/tests/unigraf/unigraf_connectivity.c
> @@ -108,10 +108,20 @@ int igt_main()
>   
>   		/* i = 0 is SST so we need to process max_count + 1 streams */
>   		for (int i = 0; i <= max_count; i++) {
> +			if (i == 1) {
> +				/* Full physical HPD disconnect/reset between SST and MST iterations */
> +				unigraf_unplug();
> +				unigraf_hpd_deassert();
> +				sleep(igt_default_display_detect_timeout());
> +				unigraf_reset();
> +				sleep(igt_default_display_detect_timeout());
> +			}

There is already a unplug/plug sequence just bellow, what is the 
difference? The reset function don't do a lot of stuff except also 
disconnecting the unigraf.

> +			unigraf_unplug();
>   			unigraf_hpd_deassert();
>   			/* Let the hardware detect the new state  */
>   			sleep(igt_default_display_detect_timeout());
>   
> +			unigraf_plug();
>   			unigraf_set_mst_stream_count(max(i, 1));
>   			if (!i)
>   				unigraf_set_sst();
> @@ -125,13 +135,18 @@ int igt_main()
>   			newly_connected_count = kms_wait_for_new_connectors(&newly_connected,
>   									    already_connected,
>   									    already_connected_count,
> +									    max(i, 1),
>   									    drm_fd);
>   
>   			diff_len = get_array_diff(newly_connected, newly_connected_count,
>   						  already_connected, already_connected_count, &diff);
>   
> -			igt_assert_f(diff_len == max(i, 1),
> -				     "Invalid connected connector count, expected %d found %d\n",
> +			/*
> +			 * We allow diff_len to be between 1 and max(i, 1) because the DRM driver may incrementally
> +			 * expose MST stream nodes rather than surfacing the entire topology at once.
> +			 */
> +			igt_assert_f(diff_len >= 1 && diff_len <= max(i, 1),
> +				     "Invalid connected connector count, expected between 1 and %d found %d\n",
>   				     max(i, 1), diff_len);

This change is a bit unexpected, the goal of this test is to properly 
check if all connectors are detected. If one is not detected, it is 
either a kernel bug or a hardware bug.

If the issue is the DRM slowness, can you just try to increase the timeout?

>   		}
>   
> diff --git a/tests/unigraf/unigraf_lt.c b/tests/unigraf/unigraf_lt.c
> index 81e6ecd3f..b5989503d 100644
> --- a/tests/unigraf/unigraf_lt.c
> +++ b/tests/unigraf/unigraf_lt.c
> @@ -107,11 +107,15 @@ int igt_main()
>   
>   		for (i = 0; i < ARRAY_SIZE(lane_counts); i++) {
>   			igt_dynamic_f("unigraf-dp-lane-count-%d", lane_counts[i]) {
> +				/* Hardware PHY requires physical settle time before and after link constraints */
> +				sleep(igt_default_display_detect_timeout());
>   				unigraf_reset();
>   				unigraf_set_max_lane_count(lane_counts[i]);
>   				unigraf_hpd_pulse(500000);
> +				sleep(igt_default_display_detect_timeout());
>   
> -				igt_display_require_output(&display);
> +				igt_display_reset_outputs(&display);
> +				connector = unigraf_get_connector(drm_fd);
>   				output = igt_output_from_connector(&display, connector);
>   				igt_assert(output);
>   				igt_dp_force_link_retrain(drm_fd, output, 2);
> @@ -119,6 +123,9 @@ int igt_main()
>   
>   				current_lanes = igt_dp_get_current_lane_count(drm_fd, output);
>   				igt_assert_eq(current_lanes, lane_counts[i]);
> +
> +				igt_modeset_disable_all_outputs(&display);
> +				igt_display_commit2(&display, COMMIT_ATOMIC);
>   			}
>   		}
>   	}
> @@ -134,14 +141,21 @@ int igt_main()
>   
>   		for (i = 0; i < ARRAY_SIZE(rates); i++) {
>   			igt_dynamic_f("unigraf-dp-link-rate-%d", rates[i]) {
> -				unigraf_reset();
> -				unigraf_set_max_link_rate(rates[i]);
> -				unigraf_hpd_pulse(1000000);
> -				igt_display_require_output(&display);
> -				igt_display_reset(&display);
> -				igt_display_require_output(&display);
> +				connector = unigraf_get_connector(drm_fd);
>   				output = igt_output_from_connector(&display, connector);
>   				igt_assert(output);
> +				
> +				/* Fast-fail if graphic card does not support enforcing this link rate natively */
> +				igt_require(igt_dp_is_link_rate_supported(drm_fd, output, unigraf_rate_to_kbs(rates[i])));
> +
> +				/* Hardware PHY requires physical settle time before and after link constraints */
> +				sleep(igt_default_display_detect_timeout());

Is this sleep before reset/pulse required?

> +				unigraf_reset();
> +				unigraf_set_max_link_rate(rates[i]);
> +				unigraf_hpd_pulse(500000);
> +				sleep(igt_default_display_detect_timeout());
> +
> +				igt_display_reset_outputs(&display);
>   				igt_dp_force_link_retrain(drm_fd, output, 2);
>   
>   				init_output_and_display_pattern(&display, output);
> @@ -150,6 +164,9 @@ int igt_main()
>   				max_supported_rate = igt_dp_get_max_supported_rate(drm_fd, output);
>   				igt_require(max_supported_rate >= unigraf_rate_to_kbs(rates[i]));
>   				igt_assert_eq(current_rate, unigraf_rate_to_kbs(rates[i]));
> +
> +				igt_modeset_disable_all_outputs(&display);
> +				igt_display_commit2(&display, COMMIT_ATOMIC);
>   			}
>   		}
>   	}


  parent reply	other threads:[~2026-07-31 17:49 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-07-15 19:29 [PATCH 1/3] android: Implement lightweight GKeyFile INI string parser in glib shim Mark Yacoub
2026-07-15 19:29 ` [PATCH 2/3] tests/unigraf: Fix DRM master and heap memory leaks in tests Mark Yacoub
2026-07-16 14:54   ` Louis Chauvet
2026-07-15 19:29 ` [PATCH 3/3] tests/unigraf: Enhance link rate support checking and hardware retrain timing Mark Yacoub
2026-07-17  8:13   ` Louis Chauvet
2026-07-20 20:53   ` [PATCH v2] " Mark Yacoub
2026-07-24 14:06     ` Mark Yacoub
2026-07-31 17:48     ` Louis Chauvet [this message]
2026-07-15 20:27 ` ✓ Xe.CI.BAT: success for series starting with [1/3] android: Implement lightweight GKeyFile INI string parser in glib shim Patchwork
2026-07-15 20:54 ` ✓ i915.CI.BAT: " Patchwork
2026-07-15 23:40 ` ✓ Xe.CI.FULL: " Patchwork
2026-07-16  3:27 ` ✓ i915.CI.Full: " Patchwork
2026-07-17  8:20 ` [PATCH 1/3] " Louis Chauvet
2026-07-17 10:18   ` Kamil Konieczny
2026-07-17 19:23     ` Louis Chauvet
2026-07-20 12:25       ` Kamil Konieczny
2026-07-17 10:27 ` Kamil Konieczny
2026-07-20 18:50 ` [PATCH v2] lib/igt_rc: Introduce generic config parser Mark Yacoub
2026-07-24 14:06   ` Mark Yacoub
2026-07-31 17:48   ` Louis Chauvet
2026-07-21  3:25 ` ✓ Xe.CI.BAT: success for series starting with [v2] lib/igt_rc: Introduce generic config parser (rev3) Patchwork
2026-07-21  3:53 ` ✓ i915.CI.BAT: " Patchwork
2026-07-21 11:48 ` ✓ Xe.CI.FULL: " Patchwork
2026-07-21 17:40 ` ✓ i915.CI.Full: " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=dfd7fa67-6120-4e60-babb-f454e181f6f7@bootlin.com \
    --to=louis.chauvet@bootlin.com \
    --cc=igt-dev@lists.freedesktop.org \
    --cc=markyacoub@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox